基于深度学习的滴灌带滴孔质量检测方法研究

Research on detection for dripping irrigation pipe holes based on deep learning

滴灌带在生产过程中需要对内镶的滴片进行打孔操作,滴孔的漏打或者打偏,在使用时会影响农作物的生长,目前还没有高效的在线滴孔质量检测的方法。文章对深度学习模型卷积神经网络和用于目标检测的YOLO算法进行了研究,提出了一种基于深度学习的滴孔质量快速检测算法。该方法首先采用YOLO算法对滴槽和滴孔定位,获取二者的坐标位置,然后获取二者的中心坐标差值,通过与预设值进行比较,判断滴孔是否合格。并将该算法应用于滴灌带生产线上进行实验分析,结果表明该方法在滴孔质量检测应用上达到了良好的效果。

In the production process of the dripping irrigation pipe,the dripping heads embeded in the pipes need to be punched holes.No holes or deflected holes can cause the deaths of cropper locally.At present,there is no effective method to detect the state of the holes online.A research on the deep learning model convolutional neural networks and YOLO algorithm for target detection is made in paper,and the paper proposes an online detection algorithm of holes based on deep learning.The algorithm first uses the YOLO network to locate the drip trough and the hole,and obtains both the coordinates of them.Then the difference between the two coordinates is calculated and compared with the predefined thresholds to judge whether the hole is qualified.Finally,the algorithm was applied to the production line of drip irrigation pipe for experimental analysis.The experiments shows good results and practicability of this method.